The White Picket Fence Dream: Effects of Assets on the Choice of Family Union Arif Mamun * January 2006 Abstract A recent strand of literature in demography argues that young unmarried Americans value marriage so highly that it is perceived as a family status to be chosen after certain economic preconditions are fulfilled – after they have achieved the so-called “white picket fence dream” (a house, surplus income etc.). Motivated by these claims, in this paper we use data from the National Longitudinal Survey of Youth 1979 to examine whether there is any direct relationship between the individual’s housing and financial assets and his/her transition into marriage or cohabitation. For both men and women, analysis using a proportional hazard model indicates a positive association of asset ownership with transition into marriage, but not with transition into cohabitation. Considering the potential endogeneity of asset accumulation with respect to the choice of family status, we implement instrumental variables probit estimation. These estimates either remove the statistical significance of the association between asset ownership and family union transitions, or identify effects that are in the opposite direction to those derived from the time-to-event analysis. JEL Classification: J12, D1 Keywords: asset, home-ownership, family, marriage, cohabitation * Researcher, Mathematica Policy Research, Inc. 600 Maryland Ave SW, Washington, DC 20024. E-mail: [email protected]I would like to thank Bob Plotnick, Claus Pörtner, George Masnick and the participants of the labor and development economics brownbag at the University of Washington for helpful comments. I am indebted to Shelly Lundberg for her guidance and support all through the development of this paper. Financial support from the Grover and Creta Ensley Fellowship in Economic Policy, and the Center for Studies in Demography and Ecology at the University of Washington is gratefully acknowledged.
Transcript
The White Picket Fence Dream:
Effects of Assets on the Choice of Family Union
Arif Mamun ∗
January 2006
Abstract
A recent strand of literature in demography argues that young unmarried Americans value marriage so highly that it is perceived as a family status to be chosen after certain economic preconditions are fulfilled – after they have achieved the so-called “white picket fence dream” (a house, surplus income etc.). Motivated by these claims, in this paper we use data from the National Longitudinal Survey of Youth 1979 to examine whether there is any direct relationship between the individual’s housing and financial assets and his/her transition into marriage or cohabitation. For both men and women, analysis using a proportional hazard model indicates a positive association of asset ownership with transition into marriage, but not with transition into cohabitation. Considering the potential endogeneity of asset accumulation with respect to the choice of family status, we implement instrumental variables probit estimation. These estimates either remove the statistical significance of the association between asset ownership and family union transitions, or identify effects that are in the opposite direction to those derived from the time-to-event analysis.
∗ Researcher, Mathematica Policy Research, Inc. 600 Maryland Ave SW, Washington, DC 20024. E-mail: [email protected] I would like to thank Bob Plotnick, Claus Pörtner, George Masnick and the participants of the labor and development economics brownbag at the University of Washington for helpful comments. I am indebted to Shelly Lundberg for her guidance and support all through the development of this paper. Financial support from the Grover and Creta Ensley Fellowship in Economic Policy, and the Center for Studies in Demography and Ecology at the University of Washington is gratefully acknowledged.
1
The White Picket Fence Dream:
Effects of Assets on the Choice of Family Union
I. Introduction
There have been dramatic changes in the family formation behavior of young American men
and women over the last four decades. During this period, the prevalence of cohabitation has been
increasing sharply while age at first marriage has also been rising, and the percentage of marriages
preceded by cohabitation has been growing substantially (Fields and Casper, 2001; Casper and
Cohen, 2000; Bumpass and Sweet 1989). These demographic changes have prompted serious
concerns from researchers as well as policy makers about a retreat from the traditional pattern of
family formation. However, both quantitative and qualitative research (e.g., Edin, 2000; Tucker,
2000; Thornton and Young-DeMarco, 2001; Gibson, Edin and McLanahan, 2003, Cherlin, 2004)
indicate that among the unmarried population there is not a large-scale lack of respect for marriage
and the traditional ways of family. Young Americans do place a high value on marriage and consider
it as part of their future. In fact, Gibson, Edin and McLanahan (2003) argue that at least for some
young unwed parents, high marital expectations may be precluding them from marrying. In a
qualitative analysis of 75 unmarried young couples in the Fragile Families study, they find that
marriage signals the “arrival” of the couple, both financially and emotionally. Because marriage is
valued so highly, it is perceived as a family status to be chosen after certain economic and relational
preconditions are fulfilled – after they have achieved the so-called “white picket fence dream”.
Similar observations are also reported in Edin (2000), drawing on qualitative interviews with 292
2
low-income single mothers in three U.S. cities. While these are interesting observations, there has
not been any attempt to substantiate the qualitative evidence using quantitative methods in a large
scale dataset. If people postpone their marriage until they can achieve the different level of living
standard that they associate with marriage – a house, surplus income etc. – one may expect to detect
a direct relationship between the individual’s housing and financial assets, and his/her transition into
marriage. In this paper, we examine this intriguing issue using data from the National Longitudinal
Survey of Youth 1979. In addition to considering transition into marriage, we also analyze the effect
of asset ownership on transition into cohabitation.
Since we are interested in identifying whether asset ownership status explains the time until
the formation of a family union, we utilize the proportional hazard model – a natural estimation
technique for analyzing time-to-event data. The time-to-event analysis would ensure that asset
ownership status prior to a family union transition is sequentially exogenous to such a transition
decision. However, the individual’s intention to form a family next period or period after may
influence his/her asset accumulation behavior in the current period. As a result, unobserved
individual heterogeneity as well as the shocks that affect the choice of family status could be
correlated with asset ownership status, and the proportional hazard model estimates would be
potentially inconsistent. To address the potential endogeneity of assets we implement instrumental
variables estimation, where the set of excluded instruments contains the interaction of monthly
averages of the 30-year fixed-rate mortgages, the 1-year adjustable-rate mortgages, the federal funds
rate, and NASDAQ stock price index, with individual’s age, ethnicity and region of residence.
In our attempt to understand the role of assets in family union choice, we consider
cohabitation and marriage comparably. Thus, in addition to bridging the void in the literature on
3
identifying the effects of assets on family union, this paper would enable us to better understand the
differences and similarities between marriage and cohabitation. Furthermore, apart from providing
insights into the mechanism underlying the individual’s decision to form families, the qualitative
evidence on assets as determinants of family union transitions are also suggestive of the potential
influence that asset building policies might have on family union behavior.1 Particularly, public
policy interventions for increasing home ownership, and enhancing savings among the low-income
group through individual development accounts, would be expected to affect the family union
choices of the participants. The results available from the current study will enable us to make more
knowledgeable estimates about whether asset-building policies could potentially influence family
union behavior.
The following section provides a conceptual discussion of the determinants of family union
decisions, and a brief review of the relevant empirical literature. Section III describes the estimation
procedures, and section IV delineates the NLSY data and summary statistics. Empirical results are
presented in Section V. Summary and conclusions follow in section VI.
II. Economic Resources and Family Union Formation: Theoretical
Perspectives and Empirical Research
II. 1. Economic resources and marriage
Economic analysis of household formation, built on the foundation of Becker’s (1973, 1974,
1991) seminal work, emphasizes the effects of economic resources on the likelihood of marriage.
1 Stern (2001), Sherraden (2001), and Seidman (2001) discuss asset-building policies in the US.
4
From a microeconomic perspective, the effect of a change in the individual’s economic prospects,
such as a rise in the wage rate, on the timing of marriage can be analyzed as the effect of a change in
the wage rate on the allocation of time among family, schooling, and market work. From this point
of view, a rise in the wage rate tends to increase market work, and in turn has a negative substitution
effect on the demand for family or schooling (Becker, 1973). However, rising wages also has an
income effect, which makes marriage and family more affordable and possibly increases the rate of
return to schooling. Hence, the marriage effect of a change in young men or women’s labor market
conditions is an empirical question.
Wilson (1987, 1996) provides one of the most widely noticed discussions of a causal link
between economic resources and the postponement of marriage among the young Americans,
particularly among the low-income black population. Working his way through the complexity of
jointly determined variables, he argues that decreased employment opportunities for men is an
exogenous “prime mover”, and determines its effect on marriage rate to be negative. One way in
which the Wilson hypothesis can be put in the context of the economic theory of marriage is to
consider marriage as the matching outcome from a search process in the marriage market where
women only consider marriage with men who have demonstrated a minimum ability to perform in
the labor market (Wood, 1995).
The theoretical discussions point to three economic factors as being potentially important:
employment and earnings of men, employment and earnings of women, and welfare benefits
(Ellwood and Crane, 1990). In addition to these three factors, other measures of economic resources
that feature prominently in the empirical literature include men’s and women’s educational
attainment, work experience, and parental resources. Before we discuss the empirical studies
5
analyzing the family union transitions, it would be appropriate to note that a persistent limitation of
many of these studies is that they are unable to address the issue of individual heterogeneity which
can directly influence the likelihood of a transition into marriage or cohabitation, while at the same
time being correlated with the individual’s economic potential. As a result, most of the findings
indicate an associative relationship between economic resources and family union transitions, rather
than a causal one. Only a handful of more careful studies have tried to overcome this challenge.
Empirical results indicate that men’s labor market opportunities are associated with
significantly higher rates of marriage, although their quantitative effects may be small (see, for
example, Xie et. al., 2003; Oppenheimer, Kalmijn and Lim, 1997; Willis and Michael, 1994; Schultz,
1994; Mare and Winship, 1991; Ellwood and Crane, 1990; MacDonald and Rindfuss, 1981). Among
a small number of studies that have been attentive in addressing the issue of endogeneity of men’s
labor market opportunities with respect their choice of family status are: Olsen and Farkas (1990),
Wood (1995), and Black, McKinnish and Sanders (2003). Olsen and Farkas (1990) utilize a waiting-
time regression model with individual fixed effects to evaluate the effect of a government program
that guaranteed employment opportunities to disadvantaged youth on family union and fertility, and
find that employment opportunity encourages the formation of consensual unions. Black,
McKinnish and Sanders (2003) study the shocks to the coal and steel industries to measure the effect
of long-term changes in demand for low-skilled workers on welfare expenditures and family
structures. Using county level panel data for the period 1969-93, they find that the expansion of
high-wage jobs for low-skilled men increased marriage rates, and reduced the incidence of female-
headed households, thereby reducing the number of families at risk for receiving welfare benefits.
6
The findings in these more careful studies are supportive of the Wilson hypothesis that
men’s labor market opportunities are fundamental in determining the marriage rates, particularly in
the low-income group. Wood (1995), on the other hand, uses SMSA level aggregated data from US
Censuses in 1970 and 1980. The results which account for the endogeneity of marriage and
‘marriageability’, suggest that shrinking pool of high earning young black men explains little of the
decline in black marriage, and thereby, contradict the Wilson hypothesis. Thus, the empirical
evidence on the effect of changes in men’s economic status on marriage rates is mixed.
Studies focusing on women’s labor market prospects have also found mixed empirical
evidence. Some of these studies have found that better economic prospects for women are
associated with declines in marriage (e.g., Aassve, 2003; Blau, Kahn and Waldfogel, 2000;
McLanahan and Casper, 1995; Willis and Michael, 1994; Schultz, 1994; Mare and Winship, 1991),
while others find that the estimated relationship between indicators of women’s economic status and
incidence of marriage is either positive or insignificant (e.g., Xie et. al. 2003; Oppenheimer and Lew,
1995; Mare and Winship, 1991).
One aspect of family union transitions that has been ignored until recently is the transition
from cohabitation into marriage. Recent empirical endeavors in this regard find mixed evidence.
Some studies have shown positive association of men’s earnings and the transition from
cohabitation to marriage (Carlson, McLanahan and England, 2004; Brown, 2000; Sanchez, Manning
and Smock, 1998; Smock and Manning, 1997), while others have reported a significant negative
association between higher men’s earnings and cohabiters’ decision to marry (Sassler and McNally,
2003, Wu and Pollard, 2000). Previous research mostly tended to indicate that there is no significant
effect of women’s economic opportunities on transition from cohabitation to marriage (Sassler and
7
McNally, 2003; Sassler and Schoen, 1999; Clarkberg, 1999), although a recent study showed that
women’s education encourage transition to marriage among young unwed mothers (Carlson,
McLanahan and England, 2004).
II.2. Economic resources and cohabitation
With a handful of recent exceptions, the existing quantitative literature on the role of
economic resources in family formation has focused exclusively on marriage, ignoring cohabitation.
The studies that consider cohabitation include Xie et. al. (2003), Clarkberg (1999), Smock and
Manning (1997), Raley (1996), and Thornton, Axinn and Teachman (1995). Two studies have
indicated that improvement in men’s economic opportunities encourage the formation of cohabiting
unions (Clarkberg, 1999; Smock and Manning, 1997), another study found that men’s school
enrollment deters entrance into cohabitation (Thornton, Axinn and Teachman, 1995), while still
others have reported no significant effect of several measures of men’s economic potential (Xie et.
al., 2003) on the rate of transition into cohabitation. With regard to women’s economic potential,
previous studies indicate that enhancing women’s economic potential discourages the formation of
cohabiting unions (Thornton, Axinn and Teachman, 1995), while others found that women’s
economic status has no significant effect on transition into cohabitation (Xie et. al., 2003). Kravdal
(1999) relates the growth in cohabitation in Norway with the issues of ‘affordability’ of marriage. He
finds that women’s cumulated income has a positive association with the likelihood of cohabitation
among women with children, but not for women without children. The author interpreted these
results as weak evidence on marriage requiring a stronger economic underpinning than cohabitation.
8
However, since sufficient attention has not been paid to the issue of selection, one has to be very
cautious in interpreting such findings any more than reflections of positive correlations.
II.3. Asset ownership and family union formation
While income is certainly critical, wealth and assets are also important complementary
measures of an individual’s command over economic resources. The individual’s assets give us an
estimate of their economic readiness to marry in relation to their ideational value of marriage.
Individuals with higher exogenous endowments are more likely to marry because they have more to
share and can provide greater access to credit and insurance (Lam, 1988). On the other hand, higher
exogenous endowments can potentially reduce the marital surplus by way of requiring less
specialization in the household, and thereby dissuade the individual from marriage. Therefore, a
priori, the effect of assets on the choice of family union is ambiguous.
More realistically, assets are not exogenous and they reflect accumulated past income, and
savings behavior. The economic model of the determinants of marriage considers the concept of
potential wage rates, instead of actual or realized wage rates. Realized wages is as much the result of the
marriage decision as its cause. Since assets (net of inheritance) are accumulated savings from realized
earnings, they are likely to be directly affected by the marriage decisions. And even inheritance can
be influenced by the individual’s choice regarding family union. Also, there is substantial empirical
research showing how family composition affects household wealth and savings (e.g., Aizcorbe,
Kennickell and Moore 2003; Lupton and Smith 2003; Wolff, 2001; Mulder and Wagner, 2001;
Browning and Lusardi 1996; Lusardi, Gossa and Krupka 2001; Avery and Kennickell 1991). Most of
these empirical studies on savings are descriptive, and they generally identify that married couples
9
have the highest levels of wealth and lone parents the lowest with singles in between (but with quite
low levels of wealth). Taken together, these studies suggest that addressing the endogeneity of assets
is the primary challenge in analyzing their effects on the choice of family form.
III. Empirical Methodology
The primary challenge in analyzing the effect of assets on the choice of family form is to
address the issue of endogeneity of assets with respect to family status. We undertake a two-pronged
approach to address this issue. First, we utilize a time-to-event analysis approach by using
proportional hazard model. Second, we implement instrumental variables estimation.
III.1. Time-to-event analysis
The central question we are examining is whether the individual’s asset ownership status
explains his/her family union transition. The question could be put forward alternatively as whether
asset ownership can explain the time elapsed until a family union transition occurs. A natural way to
empirically estimate such effects is to apply a time-to-event analysis approach. The benefit of using a
time-to-event analysis in our context is that it ensures ‘sequential exogeneity’ 2 of assets with regard
to family status. We are looking at the effect of asset ownership prior to the event of a family union
on the probability of a union in the next period. Hence, asset ownership is not sequentially
dependent on family transition decision.
2 For a discussion on sequential exogeneity, see Wooldridge, 2002, chapter 11.
10
We consider five sets of family union transitions: a) non-partnered to first partnered union
(marriage or cohabitation); b) non-partnered to first cohabitation; c) non-partnered to first marriage;
d) cohabitation to first marriage; and e) unmarried (non-partnered or cohabiting) to first marriage.
We consider only transitions into the first marriage and first cohabitation in order to keep our
analysis simple, since subsequent family unionization is confounded by the choices on dissolving the
previous family union. Since the individual’s asset ownership status varies over time, for our
purposes we utilized a Cox proportional hazard model with time-varying covariates (see Lancaster,
1990 for details) to analyze these transitions. In this model, the instantaneous hazard rate of
transitions to family union is specified for individual i, t years until the family transition occurs, as:
0[ , ( )] ( )exp[ ( )]i x ih t x t h t x tβ=
The baseline hazard, 0 ( )h t , is a nonparametric, time-varying function; ( )ix t is a vector of regressors
that includes time-varying asset ownership indicators; and xβ is the vector of coefficients to be
estimated. We used the maximum-likelihood estimation procedure available in Stata to implement
the model (Cleves, Gould and Gutierrez, 2004).
III.2. Instrumental variables estimation
Although the time-to-event analysis ensures sequential exogeneity of asset ownership with
respect to family status, the individual’s intension to form a family in the future may influence
his/her asset accumulation behavior in the current period. Therefore, unobserved individual
heterogeneity (e.g., propensity to accumulate assets or prudence, ability to form a household etc.) as
well as the shocks that affect family status could be correlated with asset ownership status. In this
11
sense, asset ownership status may not be strictly exogenous to the family transition decision, and
hence, the proportional hazard estimates are potentially inconsistent. In other words, the time-to-
event analysis may not provide us with the true effects of asset ownership on family union
transitions. Note that the true effect of asset ownership is the effect which would result were it
possible to randomly assign asset ownership to a sample of non-partnered men and women. This
true effect may be smaller than the effect we estimate by comparing the hazard rates of family union
transition for men and women with assets to the rates of transition for men and women without
assets, precisely because of the type of endogeneity that is suggested above. Since there is no
established instrumental variables framework for hazard models, we implement instrumental
variables (IV) probit estimation in a discrete-time analogue of the (continuous time) proportional
hazard model to deal with the potential endogeneity of assets.3
For the IV estimation, instead of having a standard pooled cross-sectional limited dependent
variable, we define the dependent variable as a dichotomous indicator of whether a family union
transition occurs in the next period.4 The advantage of constructing the dependent variable in this
fashion is that we are able to retain the sequential exogeneity of the asset ownership status in the
time-to-event analysis while we address the concern about strict exogeneity of asset ownership.
3 The other way to do it may be to estimate a ‘waiting-time regression model’ (Olsen and Wolpin, 1983) by linearizing the probability density function of time to failure, and estimate using linear methods. However, Olsen and Wolpin (1983) requires imposition of an exogenously chosen upper limit on the time to remain in the non-partnered status, which is deemed to be very restrictive for our analysis. 4 To use definitions from time-to-event analysis, we use the “failure indicator” in the hazard model as the dependent variable in our discrete time analysis.
12
IV. NLSY79 Data and Summary Statistics
The National Longitudinal Survey of Youth 1979 (NLSY79) is a nationally representative US
sample of young men and women who were 14-22 years old when they were first interviewed
(CHRR, 2001). The respondents were interviewed annually until 1994, and biennially since then.
Data from the first through the 19 th (2000) round are used for this paper. We have used data from
the earliest round to determine the respondent’s family life history. Detailed information on wealth
and assets are, however, available only since the 7 th round (i.e., 1985).5
We stratify the data by gender, and all our analyses are conducted separately for men and
women. We observe the family life transitions of men and women in our sample during the period
1985 to 2000. The time-until-transition to a family union (in marriage, or in cohabitation) is
identified by combining information on current family status, beginning dates of cohabiting and
marital relationships, interview dates, as well as partner identifiers available in NLSY79. Table 1
presents some summary statistics on the unmarried spells considered here. As the table shows, the
median duration in non-partnered spell for both men and women in our sample is more than 11
years. This is not surprising given that the beginning of this spell is either the date of their first
interview in NLSY or their 14 birthday, whichever is later. The median duration in cohabitation is
about 2.7 years. Table 2 provides the number of events we observe in our data for each category of
transition, along with the median duration prior to any transition. During the period under analysis,
there are 1525 transitions into first-marriage among women; 659 of these transitions are from
cohabitation. For men, there are 1807 transitions into first-marriage of which 683 are from
cohabitation. We also have 1304 transitions for women from a non-partnered (i.e., never-married
5 Due to budgetary restrictions, wealth questions were not administered in 1991 and 2002 rounds of NLSY79.
13
non-cohabiting) status to a family union – either in marriage or in cohabitation – of which 422 are
into cohabitation. Among men, we observe 1877 transitions from a non-partnered status to a family
union, and 722 of these transitions are into cohabitation.
In Table 3, we present snapshots on the different characteristics of the respondents in our
sample. The table is intended to provide a glimpse into the nature of the sample we have for our
analysis. The table shows that the proportions that are married increased over time for both men
and women, although the fraction of men married converged to the fraction of women married only
in the later years. Summary statistics are presented for men and women in three different family
statuses: non-partnered (never-married and non-cohabiting), cohabiting and married in 1985.
As we examine the effect of asset ownership on family union transitions in this study, we
take three types of assets into consideration6: home ownership; liquid monetary assets as indicated
by the availability of funds in savings account, certificates of deposit, money market instruments and
IRA-Keoghs; and financial investments in stocks, bonds, and mutual funds. In our empirical analysis
we include dichotomous indicators of ownership of these three types of assets. While data on home
ownership and liquid monetary assets are available since 1985, stocks-bonds-mutual funds data are
available only from 1988.
Table 3 reveals that both married men and married women are significantly more likely to be
a homeowner. On average, married men and women are also more likely to own liquid monetary
assets, as well as investments in stocks, bonds and mutual funds. This is all the more clear from
Figures 1 through 6 showing the average asset ownership status of non-partnered, cohabiting and
married men and women at different age. Although the fraction that are homeowners increases with
6 These ownership data are available from the “Asset” section of the NLSY79, and the relevant questions ask about the ownership status of the respondent and their spouse/partner with respect to the particular categories of assets.
14
age for men and women in any family status, at every age those who are married are more likely to
be homeowners than those who are not married. Cohabiting men are marginally more likely to be
homeowners than non-partnered men, and the same is true for young cohabiting women. A much
larger fraction of cohabiting women in their mid-30s are homeowners compared to the non-
partnered in that age group. In terms ownership of monetary assets, married men in their early 20’s
appear to be similar to cohabiting men. However, married men who are in their late 20’s or older,
are more likely to have some monetary assets compared to men in the other two family statuses.
Married women, on the other hand, are more likely to have some monetary assets at about any age
compared to the unmarried women. The proportion owning financial investments is quite low for
both men and women in any of the three family statuses.7 However, even at that low level, married
men and women at any age are more likely to own some financial investments than their unmarried
counterparts. Interestingly, cohabiting men and women are either less (in their 20’s) or equally (in
their 30’s) likely to own monetary assets as well as financial investment compared to non-partnered
men and women.
The other characteristics that are considered in our analysis and are summarized in Table 3
include age, race and ethnicity, own education, income (wage and business income) in the past
calendar year, welfare recipiency in the past calendar year (includes receipt of cash assistance from
AFDC or TANF, supplemental security income, food stamps, housing support, or any other
benefit), religion, father’s and mother’s education, region of residence, whether or not the state of
residence recognizes common law marriage, and local unemployment rate. Men and women in our
sample are between 20 to 28 years old in 1985, the year since which we follow their family union
transitions. On average, married men and women are older than their counterparts in the other two 7 In figures 5 and 6, the outlier percentages in the highest age group reflect small sample in that age group.
15
family statuses. Married men and women are also likely to be at least high school graduates than
others. Table 3 also indicates that while women in general are more likely to be welfare recipients
than men, larger fraction of unmarried women received public assistance than their married
counterparts.
V. Results
V. 1. Results from time-to-event analysis
Estimates from the Cox proportional hazard model with the asset ownership variables as
time-varying covariates are presented in Table 4. As already noted, in analyzing the relationship
between family union transitions and asset ownership, we are considering the ownership status of
housing assets, of liquid financial assets, and of stocks, bonds and mutual funds. Since data on
stocks and bonds ownership is available from a later period (from 1988, instead of 1985), in our
analysis we estimate two sets of specifications – one that excludes stocks-bonds-mutual funds
ownership indicator (Model 1), and another that includes it (Model 2). Table 4 presents the
estimated hazard ratios only for the covariates of interest, and estimates for the complete
specification are available upon request. We discuss the estimated results for men and women
separately.
Results in section (a) in Table 4 show that home ownership and money in the savings
account have no statistically significant effect on women’s transition from never-married status into
a partnered relationship (either in marriage or in cohabitation). However, as the ownership of stocks
and bonds is included in the specification (Model 2), thereby reducing the period of analysis and the
16
sample size, the estimated hazard rate on women’s home ownership becomes significantly negative,
while the hazard rate on monetary assets becomes significantly positive. Table A-1 in the Appendix
reports an additional set of estimates where measures of wage and business income in the previous
calendar year, and an indicator of the individual’s public assistance recipiency (AFDC/TANF, food
stamps, unemployment insurance benefits, supplemental security income etc.) in the previous
calendar year is added to the specifications in Table 4. While we wanted to examine whether
inclusion of income and welfare recipiency in the specification have any major influence on the
estimated effects of asset ownership, their potential endogeneity with respect to family union
transitions convinced us for not including them in our initial specifications. As it appears, inclusion
of these two variables does not change the estimated effects of asset ownership on transition to a
family union in any substantive way. In addition, welfare recipiency tends to be negatively correlated
with transition into a family union, a result that resonates with a large existing literature (see Moffitt,
1998 for a recent review).
Next, we perform a competing-risks analysis that allows for two ways to exit the non-
partnered status: form a family in cohabitation, or in marriage. Estimated hazard ratios for
transitions from non-partnered to cohabitation are reported in section (b) in Table 4. For women,
both Model 1 and Model 2 show that ownership of home as well as liquid monetary assets are not
significantly associated with forming a cohabiting relationship. Results from section (c) show that
while ownership of liquid assets is positively associated with the rate of marriage for non-partnered
women, home ownership in this group has a negative association with transition into marriage. As it
appears, the pattern of association of home as well as liquid asset ownership with women’s rate of
forming any family union (in section a) is primarily driven by the association between ownership of
17
these assets and transition into marriage. The positive association between liquid asset ownership
and marital transitions is also emphasized by the estimates in section (d) which indicate that women
who have access to liquid assets are more likely to marry their cohabiting partners. Results in section
(d) further show that home ownership has no significant relationship with the rate of marriage
among cohabiting women. While the results in this section for liquid assets conform to the
qualitative evidence that women would chose to marry their cohabiting partners when there is
surplus income (e.g., Gibson, Edin and McLanahan, 2003; Cherlin, 2004), the insignificant
relationship between home ownership and marital transition for cohabiting women does not
correspond to a so-called ‘white picket fence’ explanation. As we consider women’s transition into
marriage from either a never-married or a cohabiting status (section e, Table 4), we find that both
home and liquid assets ownership have positive correlation with such transitions.
Overall, results from Cox proportional hazard model estimates suggest that liquid assets are
positively associated with women’s rate of transition into marriage. For home ownership, there is
weak evidence of a negative relationship. In the case of transition into cohabitation, asset ownership
does not have a statistically significant association with women’s choice of cohabitation.
For men, the overall evidence indicates that asset ownership is positively associated with rate
of transition to marriage, but not to cohabitation. Specifically, the rate of marital transition by non-
partnered men has a significant positive relationship with both home and liquid asset ownership, but
no significant relationship exist for men’s transition from non-partnered status to cohabitation (see
sections a, b, c, and e for men in Table 4). More interestingly, cohabiting men’s rate of transition to
marriage has no significant relationship with home ownership, but there is positive significant
association with ownership of liquid assets and financial investments. These results give the
18
impression that while ownership of all types of assets is positively correlated with the never-married
men’s decision to marry; only liquid assets are significantly correlated with cohabiting men’s choice
to marry.
It would be appropriate to note that the individual’s race and ethnicity indicators have been
included in all the specifications reported, along with the other control variables. Overall, for all the
transitions analyzed here, compared to non-black non-Hispanic women, black women are less likely
to be in any type family union – in marriage or in cohabitation. Black men, on the other hand, are
significantly more likely to transit to a cohabiting relationship, and less likely to transit into marriage
in comparison with their non-black non-Hispanic counterparts. Our estimates for Hispanic women
are similar to those for Black women, but somewhat weaker. For Hispanic men, we have some
evidence that unlike black men, they are less likely either to marry or to cohabit than non-black non-
Hispanic men.
The empirical estimates discussed so far provide some interesting new evidence. Taken
together they show that home ownership as well as access to liquid assets is positively associated
with rates of marital transition from a non-partnered status, particularly for men. There is weak
evidence of a negative association between home ownership and women’s marital transitions. Home
ownership is not associated with rates of marital transition among cohabiting men and women in the
sample, although access to liquid monetary assets is. In addition, both for men and women, asset
ownership is not associated with the rate of transition into cohabitation. These results only partially
agree with the previously discussed qualitative evidence that suggests a stronger relationship,
particularly between home ownership and rates of transition from cohabitation to marriage.
Moreover, the results for the transition from cohabitation to marriage might be considered as
19
suggestive evidence that both men and women would accumulate monetary assets while cohabiting,
and buy a house when their commitment to a relationship is sealed by the decision to marry. This
only reinforces the apprehension about the endogeneity of asset accumulation that we have noted
earlier.
V. 2. Results from instrumental variables estimation
To address the potential endogeneity of assets we implement instrumental variables (IV)
estimation. To operationalize the procedure, we convert the structure of analysis from a continuous
time to a discrete time hazard model with the dependent variable being a dichotomous indicator of a
family union transition in the next period. As we treat the asset ownership status as endogenous to
family union transitions, the set of excluded instruments is constructed as follows. We use the
interaction of monthly averages of the 30-year fixed-rate mortgages, the 1-year adjustable-rate
mortgages,8 the federal funds rate, and NASDAQ stock price index, with individual’s age, ethnicity
and region of residence as the set of excluded instruments.9 The individual’s homeownership
decision is expected to be influenced by the mortgage rates as they play a dual role in the housing
market: on the one hand, mortgage rates show the time-value-of money; and on the other hand, they
have key roles in determining housing prices. The stock price index and the federal funds rate are
proxies for the return to savings and investment during the sample period, and are therefore
expected to influence the individual’s stock of assets at any point in time. Interacting the mortgage
8 The monthly average mortgage rates are collected from the Freddie Mac Survey of Commitment Points and Rates. 9 Housing price could not be used as an instrument since it directly affects the decision to form independent households (e.g., see, Borsch -Supan, 1986, Haurin et. al, 1993).
20
rates, stock price index and the federal funds rate with the individual characteristics provides us with
the individual level variation required for identification of the estimating equation.
Test statistics on the joint-significance of the excluded instruments in the first-stage
regressions are provided in Table 5. Although the F-statistics for ownership of stocks and bonds is
quite small, the values of the F-statistic on the other two asset ownership indicators are sufficiently
high to remove the greater portion, if not all, of the potential bias in an analysis conducted without
regard to the endogeneity of assets (Bound et. al. 1995, Table A-1; Hahn and Hausman, 2002). The
validity of these instruments, particularly for the specifications that includes only home and liquid
asset ownership, is also underlined by the Hansen-Sargan J statistic for over-identification tests
reported in Table 6 along with the IV-probit estimates.
Table 6 reports both probit and IV-probit estimates for specifications described earlier as
Model 1 and Model 2. We applied Newey’s (1987) efficient two-step estimator in Stata to estimate
the IV-probit specifications. The sign and significance of the estimated coefficients on asset
indicator variables in the probit model are similar to those in the Cox proportional hazard model for
any family union transitions considered, and for both men and women.
IV-probit estimates for women show that addressing the potential endogeneity of asset
ownership removes the positive association between asset ownership and family union transitions.
On the contrary, access to monetary assets has a negative influence on the rates of transition to
marriage for non-partnered women (sections a, c and e in Table 6). Inclusion of income and welfare
recipiency in the specification (not reported in the table) does not change these results in any
important way. Section (b) in Table 6 indicates that asset ownership has no statistically significant
influence on women’s rate of transition to cohabitation as we account for the endogeneity of assets.
21
IV probit estimates in section (d) of Table 6 show that home ownership reduces women’s likelihood
of marrying their cohabiting partner, although such negative effect does not remain statistically
significant when we include ownership of stocks and bonds in our specification.
For men, IV probit estimates show that the effects of asset ownership on the rates of
transition into cohabitation as well as transition into marriage, from either non-partnered or
cohabiting status, are not significantly different from zero. For men’s transitions into marriage from
any unmarried status (section e in Table 6), neither homeownership nor monetary asset-ownership
has any significant effect. However, as we add the indicator for ownership of stocks and bonds in
the specification, it has a statistically significant negative effect on the rate of marital transition for
men.
All together, the IV probit estimates either remove the statistical significance of the
association between asset ownership and family union transitions, or indicate effects that are in the
opposite direction to those derived from the time-to-event analysis. However, it has been suggested
in the literature that the influence of economic factors in family life transitions might be different for
different education groups (e.g., Moffitt, 2000). To examine whether our results also reflect a similar
pattern, we conducted the instrumental variables analysis by stratifying our sample into two groups:
those who have never attended college (i.e., at most high-school graduate), and those who have
some college or more education (i.e., more than high school education). In the stratified sample, we
implemented the IV-probit estimation for Model 1 only. Further, the F statistics for the joint
significance of the instruments in the first stage estimation indicate that as we stratify the sample, the
instruments perform satisfactorily only in the case of transitions from unmarried (never-
married/cohabiting) to married status. In Table 7 we report the IV-probit results for this particular
22
transition, for men and women disaggregated by education groups. A comparison of these estimates
with those reported in section (e), Model 1 in Table 6 suggests that the estimated effects of asset
ownership on marital transitions do not vary significantly between the two education-groups, for
both men and women. While these auxiliary results are generated only for the transitions from
unmarried to married status, they provide enhanced confidence in our IV estimates for all the other
transitions identified across all education levels.
VI. Conclusion
Motivated by a recent set of findings by demographers, the paper presents two broad sets of
evidence on the relationship between asset ownership and the family union transition decisions by
men and women. The first set of findings, from Cox proportional hazard model with time varying
covariates, reveal that home ownership as well as access to liquid assets is positively associated with
rates of marital transition from a non-partnered status, particularly so for men. However, home
ownership is not associated with rates of marital transition among cohabiting men and women in the
sample, although access to liquid monetary assets is positively associated with rates of cohabiting
men and women marrying their partners. Moreover, the rates of transition into cohabitation among
men and women are not at all related to their asset ownership status.
The second set of findings stem from instrumental variables probit estimation, implemented
to remove the potential bias in the hazard model estimates. The bias is anticipated due to the
endogeneity of asset accumulation with respect to the family union decisions. The IV probit
estimates either remove the statistical significance of the association between asset ownership and
family union transitions, or indicate effects that are in the opposite direction to those derived from
23
the estimated hazard model. As with any instrumental variables estimation, the strength of the set of
identifying variables in playing the role of instruments is crucial. While we would not claim that
these results are definitive, we do believe that they are suggestive. In relating asset ownership to
family union behavior, researchers need to be aware that it is likely that common preferences or
opportunities underlie both the decisions to accumulate asset and the decision to form a family by
marriage or by cohabitation.
Results from the time-to-event analysis indicate that at least as a behavioral regularity we
observe a positive relationship between asset ownership and marital transitions. The IV estimates,
however, suggest that such behavioral regularity does not indicate a causal relationship. In other
words, those who are inherently more likely to marry are the ones who would accumulate assets, and
hence asset ownership does not cause their transition into marriage. Indeed, homeownership and
monetary assets may reduce the marital surplus by reducing the level of specialization in the
household, and thereby influence people to delay marriage, as suggested by the IV results. More
importantly, these results have notable policy implications. If the IV estimates are reflective of the
true effects of asset ownership, provision of housing subsidy or incentive to accumulate assets may
not lead to any significant improvement in the rates of marriage.
24
References
Aassve, Arnstein. 2003. “The Impact of Economic Resources on Premarital Childbearing and Subsequent Marriage Among Young American Women”. Demography. 40(1). pp. 105-26. Aizcorbe, Ana and Kennickell Arthur and Kevin Moore. 2003. “Recent changes in U.S. family finances: Evidence from the 1998 and 2001 Survey of Consumer Finance”. Federal Reserve Bulletin. January. 89(1). Avery, Robert and Kennickell, Arthur. 1991. “Household Savings in the US”. Review of Income and Wealth. December. 37(4). pp. 409-32. Becker, Gary S. 1973. “A Theory of Marriage: Part I”. Journal of Political Economy. 81(4). pp. 813-46. Becker, Gary S. 1974. “A Theory of Marriage: Part II”. Journal of Political Economy. 82(2-II). pp. S11-S26. Becker, Gary S. 1991. A Treatise on the Family. Enlarged Edition. Harvard University Press. Cambridge, MA. Black, Dan; McKinnish, Terra; and Sanders, Seth. 2003. “Does the Availability of High-Wage Jobs for Low-Skilled Men Affect Welfare Expenditures? Evidence from Shocks to the Steel and Coal Industries”. Journal of Public Economics. 87. pp. 1921- 42. Blau, Francine; Kahn, Lawrence; and Waldfogel, Jean. 2000. “Understanding Young Women’s Marriage Decisions: The Role of Labor and Marriage Market Conditions”. Industrial and Labor Relations Review. July. 53(4). pp. 624-47. Borsch-Supan, Axel. 1986. “Household Formation, Housing Prices, and Public Policy Impacts”. Journal of Public Economics. 30. pp. 145-64. Bound, John; Jaeger, David and Baker, Regina. 1995. “Problems with Instrumental Variables Estimation When the Correlation Between the Instruments and the Endogenous Explanatory Variable is Weak”. Journal of American Statistical Association. 90(430). pp. 443-50. Bowman, Cynthia G. 2004. “Legal Treatment of Cohabitation in the United States”. Law & Policy. January. 26(1). pp. 119-51. Brown, S., 2000. “Union Transitions among Cohabitors: The Significance of Relationship Assessments and Expectations. Journal of Marriage and the Family. 62 (3). pp. 833–47. Browning, Martin and Annamaria Lusardi. 1996. “Household Saving: Micro Theories and Micro Facts”. Journal of Economics Literature. 34(December). pp1797-1855 Bumpass, Larry L. and Hsien-Hen Lu. 2000. “Trends in Cohabitation and Implications for Children’s Family Contexts in the United States”. Population Studies 54. pp. 29-41. Bumpass, Larry L. and James A. Sweet. 1989. “National Estimates of Cohabitation”. Demography. 26. pp. 615-25.
25
Carlson, Marcia; McLanahan, Sarah and England, Paula. 2004. “Union Formation in Fragile Families”. Demography. 41(2). May. pp. 237-61. Casper, Lynne M. and Cohen, Philip. N. 2000. “How does POSSLQ Measure Up? Historical Estimates of Cohabitation”. Demography. 37(2). pp. 237-45. Center for Human Resource Research (CHRR). 2001. NLSY79 User’s Guide: A Guide to the 1979-2000 National Longitudinal Survey of Youth Data. The Ohio State University. Columbus, Ohio. Cherlin, Andrew. 2004. “The Deinstitutionalization of American Marriage”. Journal of Marriage and Family. 66. November. pp. 848-61. Clarkberg, M., 1999. The Price of Partnering: The Role of Economic Well-Being in Young Adults’ First Union Experiences. Social Forces. 77 (3). pp. 945–68. Cleves, Mario, William Gould and Gutierrez, Roberto. 2004. An Introduction to Survival Analysis Using Stata (Revised Edition). Stata Corporation. College Station, TX Easterlin, Richard A. “Relative Economic Status and the American Fertility Swing”. In E.B. Sheldon (ed.) Family Economic Behaviour: Problems and Prospects. Lippincott. Toronto and Philadelphia. Edin, Kathryn. 2000. “What Do Low-Income Single Mothers Say About Marriage?”. Social Problems. 47. pp. 112-33. Ellwood, David and Crane, Jonathan. 1990. “Family Change among Black Americans: What Do We Know?”. Journal of Economic Perspective. 4(4). pp. 65-84. Fields, Jason and Casper, Lynne M.. 2001. “America’s Families and Living Arrangements: March 2000”. Current Population Reports, P20-537. U.S. Census Bureau, Washington, DC. Gibson, Christina; Kathryn Edin and Sara McLanahan. “High Hopes But Even Higher Expectations: The Retreat From Marriage Among Low-Income Couple”. Center for Research on Child Wellbeing Working Paper # 03-06-FF. June 2003 (Revised: June 2004). Hahn, Jinyong and Hausman, Jerry. 2002. “Notes on bias in estimators for simultaneous equation models”. Economics Letters. 75. pp. 237–241. Haurin, Donald; Hendershott, Patric; and Kim, Dongwook, 1993. “The Impact of Real Rents and Wages on Household Formation”. The Review of Economics and Statistics. 75(2). pp. 284-93. Lam, David. 1988. “Marriage Markets and Assortative Mating with Household Public Goods: Theoretical Results and Empirical Implicaitons”. Journal of Human Resources. 23(4). pp. 462-87. Lancaster, Tony. 1990. The Econometric Analysis of Transition Data. Cambridge University Press. Cambridge, UK. Lupton, Joseph and Smith, James. 2003. “Marriage, Assets and Saving”. Chapter 6 in Marriage and the Economy: Theory and Evidence from Advanced Industrial Societies. by Shoshana Grossbard-Shechtman (ed). Cambridge University Press.
26
Lusardi, Annamaria; Cossa, Ricardo and Krupka, Erin. 2001. “Savings of Young Parents”. Journal of Human Resources. (Special Issue on Early Results from the National Longitudinal Survey of Youth 1997). Autumn. 36(4). pp. 762-94. MacDonald, Maurice M. and Rindfuss, Ronald R. 1981. “Earnings, Relative Income, and Family Formation”. Demography. 18(2). pp. 123-36. Manning, W.D., Smock, P.J., 1995. “Why Marry? Race and the Transition to Marriage among Cohabitors” Demography. 32 (4). pp. 509–20. Mare, R. D. and Winship, C. 1991. “Socioeconomic Change and the Decline of Marriage for Blacks and Whites”. pp. 175-202 in The Urban Underclass. edited by C. Jencks and P.E. Peterson. The Urban Institute. Washington, DC. McLanahan, Sara and Casper, Lynne. 1995. “Growing Diversity and Inequality in the American Family”, In State of the Union: America in the 1990s. Edited by Reynolds Farley. Vol. 2. pp. 1-45. Russell Sage Foundation. New York. Moffitt, Robert A. 1998. “The Effect of Welfare on Marriage and Fertility”. In Robert Moffitt edited Welfare, the Family, and Reproductive Behavior. pp. 50-97. National Academy Press. Washington, DC. Moffitt, Robert A. 2000. “Female Wages, Male Wages, and the Economic Model of Marriage: The Basic Evidence”. In Ties That Bind: Perspectives on Marriage and Cohabitaiton. Edited by Linda Waite, Christine Bachrach, Michelle Hindin, Elizabeth Thomson, and Arland Thornton. Aldine de Gruyter. New York. Mulder, Clara and Wagner, Michael. 2001. “The Connections between Family Formation and First-time Home Ownership in the Context of West Germany and the Netherlands”. European Journal of Population. 17(2). pp. 137-64. Newey, Whitney K. 1987. “Efficient Estimation of Limited Dependent Variable Models with Endogenous Explanatory Variables”. Journal of Econometrics. 36. pp. 231-50. Olsen, Randall and Wolpin, Kenneth. 1983. “The Impact of Exogenous Child Mortality on Fertility: A Waiting Time Regression with Dynamic Regressors”. Econometrica. 51. pp. 731-49. Olsen, R., Farkas, G., 1990. The effect of economic opportunity and family background on adolescent cohabitation and childbearing among low-income blacks. Journal of Labor Economics. 8 (3), 341–362. Oppenheimer, V. and Lew, V. 1995. “American Marriage Formation in the 1980s: How Important Was Women’s Economic Independence?”. In Gender and Family Change in Industrialized Countries. Edited by K.O. Mason and A.M. Jensen. pp. 105-38. Clarendon Press. Oxford, England. Oppenheimer, V., Kalmjin, M. and Lim, N. 1997. “Men’s Career Development and Marriage Timing During a Period of Rising Inequality”. Demography. 34. pp. 311-30. Oppenheimer, Valerie K. 1997. “Women’s Employment and the Gains to Marriage: The Specialization and Trading Model of Marriage”. Annual Review of Sociology. 23. pp. 431-53.
27
Raley, R.K., 1996. “A Shortage of Marriageable Men? A Note on the Role of Cohabitation in Black–White Differences in Marriage Rates”. American Sociological Review. 61. pp. 973–83. Rivers, Douglas and Vuong, Quang. 1988. “Limited Information Estimators and Exogeneity Tests for Simultaneous Probit Models”. Journal of Econometrics. Novemeber. 39(3). pp. 347-66. Sanchez, L., Manning, W., Smock, P. 1998. “Sex-specialized or Collaborative Mate Selection. Union Transitions among Cohabitors”. Social Science Research. 27. pp. 280–94. Sassler, S. and McNally, J. 2003. “Cohabiting Couples’ Economic Circumstances and Union Transitions: A Re-examination Using Multiple Imputation Techniques”. Social Science Research. 32. pp. 533-78. Sassler, S., Schoen, R., 1999. “The effect of Attitudes and Economic Activity on Marriage. Journal of Marriage and the Family. 61. pp. 147–59. Schultz, T. Paul. 1994. “Marital Status and Fertility in the United States: Welfare and Labor in the Market effects”. Journal of Human Resources. 29(2). Pp, 636-69. Seidman, Lawrence. “Assets and the Tax Code”. In Assets for the Poor: The Benefits of Spreading Asset Ownership. Edited by Thomas Shapiro and Edward Wolff. pp. 324-56. Russell Sage Foundation. New York. Sherraden, Michael. 2001. “Asset-Building Policy and Programs for the Poor”. In Assets for the Poor: The Benefits of Spreading Asset Ownership. Edited by Thomas Shapiro and Edward Wolff. pp. 302-23. Russell Sage Foundation. New York. Smock, Pamela J., and Wendy D. Manning. 1997. “Cohabiting Partners’ Economic Circumstances and Marriage.” Demography 34(3). pp. 331-41. Spitko, E. Gary. 2002. “An Accrual/Multi-Factor Approach to Intestate Inheritance Rights for Unmarried Committed Partners”. Oregon Law Review. 255. Summer. 81(2). Stern, Mark. 2001. “The Un(credit)worthy Poor: Historical Perspectives on Policies to Expand Assets and Credit”. In Assets for the Poor: The Benefits of Spreading Asset Ownership. Edited by Thomas Shapiro and Edward Wolff. pp. 269-301. Russell Sage Foundation. New York. Thornton, Arland and Linda Young-DeMarco. 2001. "Four Decades of Trends in Attitudes Toward Family Issues in the United States: The 1960s Through the 1990s." Journal of Marriage and the Family 63(4):1009-1037. Tucker, M. Belind. 2000. “Marital Values and Expectations in Context: Results from a 21-city Survey”. In Ties That Bind: Perspectives on Marriage and Cohabitaiton. Edited by Linda Waite, Christine Bachrach, Michelle Hindin, Elizabeth Thomson, and Arland Thornton. Aldine de Gruyter. New York. Willis, Robert and Michael, Robert. 1994. “Innovation in Family Formation: Evidence on Cohabitation in the United States”. In John Ermisch and Naohiro Ogawa (Eds). The Family, the Market and the State in Aging Societies. Oxford University Press. pp. 9-45. Wilson, William. J., 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. University of Chicago Press, Chicago.
28
Wilson, William. J., 1996. When Work Disappears: The World of the New Urban Poor. Alfred A. Knopf, New York. Wolff, Edward N. 2001. “Recent Trends in Wealth Ownership, from 1983 to 1998”. In Assets for the Poor: The Benefits of Spreading Asset Ownership. Edited by Thomas Shapiro and Edward Wolff. pp. 34-73. Russell Sage Foundation. New York. Wood, Robert. 1995. “Marriage Rates and Marriageable Men: A Test of Wilson Hypothesis”. Journal of Human Resources. 30(1). pp. 163-93. Wooldridge, Jeffrey. 2002. Econometric Analysis of Cross Section and Panel Data. The MIT Press. Cambridge, MA. Wu, Z., Pollard, M., 2000. “Economic Circumstances and the Stability of Nonmarital Cohabitation”. Journal of Family Issues. 21(3). pp. 303-28. Xie, Yu; Raymo, James; Goyette, Kimberly and Thornton, Arland. 2003. “Economic Potential and Entry into Marriage and Cohabitation”. Demography. May. 20(3). pp. 351-67.
29
Table 1. Summary Statistics on Spells in Different Family Status Prior to Union Transitions
Table 2. Number of Different Family Union Transitions
NLSY79 Women and Men; 1985-2000
Transitions No. of Transitions Conditional Median Duration (months) a
Women Non-partnered to partnered 1304 25.6 Non-partnered to cohabitation 422 24.5 Non-partnered to marriage 882 26.1 Cohabiting to marriage 659 17.7 Unmarried to marriage 1525 22.1 Men Non-partnered to partnered 1877 25.4 Non-partnered to cohabitation 722 24.4 Non-partnered to marriage 1155 25.8 Cohabiting to marriage 683 18.2 Unmarried to marriage 1807 22.9
Note: a. Duration conditional on the fact that the individual made the transition.
30
Table 3: Summary of Key Variables by Family Status in 1985 Women and Men NLSY79
Women (N=4535) Men (N=4350) 1985 Status
Variables
Never Partnered
Cohabit Married Never Partnered
Cohabit Married
Proportion in 1985 0.40 0.09 0.43 0.57 0.07 0.30 Proportion in 1990 0.20 0.08 0.57 0.28 0.10 0.50
Proportion in status a
Proportion in 2000 0.12 0.05 0.61 0.14 0.07 0.60 Own house 0.02 0.07 0.38 0.03 0.10 0.33 Have money asset 0.57 0.49 0.68 0.54 0.46 0.66
Assets
Own stocks & bonds b 0.09 0.04 0.14 0.11 0.05 0.16
Age Age (years) 22.93 23.40 24.17 22.86 23.81 24.52 Hispanic 0.15 0.14 0.18 0.15 0.23 0.18 Ethnicity Black 0.38 0.22 0.15 0.31 0.30 0.15 HS grad 0.37 0.47 0.50 0.41 0.46 0.48 Some College 0.35 0.18 0.20 0.28 0.14 0.15
Education
College grad 0.16 0.07 0.10 0.12 0.05 0.12 Income Annual income ($) 6347 6326 6256 8477 11202 14604 Welfare Welfare recipiency 0.20 0.29 0.18 0.05 0.09 0.13
Jew 0.01 0.00 0.00 0.01 0.00 0.00 HS grad 0.29 0.30 0.29 0.29 0.28 0.30 Some College 0.08 0.07 0.07 0.09 0.07 0.07 College grad 0.14 0.09 0.10 0.14 0.09 0.11
Father education
Missing 0.13 0.18 0.10 0.13 0.14 0.11 HS grad 0.35 0.35 0.38 0.38 0.36 0.38 Some College 0.10 0.09 0.07 0.09 0.07 0.08 College grad 0.09 0.04 0.05 0.08 0.05 0.06
Mother's education
Missing 0.05 0.05 0.05 0.07 0.10 0.06 North Central 0.24 0.27 0.24 0.24 0.22 0.25 South 0.40 0.33 0.41 0.35 0.29 0.39
Region of Residence
West 0.15 0.22 0.20 0.19 0.28 0.22 State Common Law
Marriage 0.34 0.28 0.35 0.33 0.32 0.37 Local Unemployment rate 8.03 8.06 8.36 8.06 7.81 8.36
Note: a. The remaining sample include divorced or widowed, and those who are single after cohabiting. b. For 1988, as data on stocks, bonds and mutual funds is available only since the 1988 round of NLSY79.
Figure 6. Percent of Men with Financial Investment by Family Status:
1988-2000 (NLSY79)
34
Table 4. Determinants of the Rate of Transitions Hazard ratios from Cox Proportional Hazard Model
Women Men Variables Model 1 Model 2 Model 1 Model 2
a. Transitions from Non-partnered to Partnered Own house 0.997 0.697 1.308 1.021 (0.03) (2.29)** (3.47)*** (0.18) Have money asset 1.171 1.333 1.313 1.481 (1.86) (2.12)** (3.25)*** (3.29)*** Own stocks & bonds 1.239 1.068 (1.75) (0.45) Hispanic 0.871 1.177 0.730 0.835 (0.92) (1.04) (1.73) (1.40) Black 0.554 0.407 0.893 0.774 (6.20)*** (5.30)*** (1.25) (1.98)** N 11073 6342 14717 8280 b. Transitions from Non-partnered to Cohabiting Own house 1.019 0.710 1.143 0.960 (0.10) (1.29) (0.95) (0.19) Have money asset 0.784 0.916 1.028 1.207 (1.74) (0.40) (0.26) (1.33) Own stocks & bonds 1.041 0.859 (0.16) (0.61) Hispanic 0.710 1.093 0.508 0.483 (1.39) (0.31) (2.80)*** (3.11)*** Black 0.571 0.439 1.343 1.227 (3.43)*** (2.95)*** (2.54)** (1.25) N 11073 6342 14717 8280 c. Transitions from Non-partnered to Married Own house 0.984 0.685 1.416 1.056 (0.12) (2.05)** (3.62)*** (0.39) Have money asset 1.424 1.639 1.567 1.770 (3.51)*** (2.92)*** (4.31)*** (3.72)*** Own stocks & bonds 1.329 1.191 (1.94) (1.16) Hispanic 0.943 1.186 0.896 1.126 (0.38) (0.88) (0.63) (0.75) Black 0.550 0.403 0.654 0.530 (5.36)*** (4.73)*** (3.54)*** (3.70)*** N 11073 6342 14717 8280
35
Table 4 (Cont’d).
Women Men Variables Model 1 Model 2 Model 1 Model 2 d. Transitions from Cohabiting to Married Own house 1.066 1.120 0.972 0.966 (0.64) (0.96) (0.29) (0.33) Have money asset 1.437 1.388 1.831 1.695 (4.15)*** (2.99)*** (7.02)*** (5.05)*** Own stocks & bonds 1.279 1.514 (1.52) (3.23)*** Hispanic 0.718 0.791 0.503 0.463 (2.56)** (1.50) (5.06)*** (4.73)*** Black 0.578 0.562 0.421 0.457 (4.27)*** (3.69)*** (7.33)*** (5.66)*** N 3668 2496 4076 3053 e. Transitions from Unmarried (Non-partnered/Cohabiting) to Married Own house 1.279 1.152 1.416 1.257 (2.96)*** (1.50) (4.84)*** (2.56)** Have money asset 1.336 1.306 1.538 1.513 (4.49)*** (2.93)*** (7.13)*** (4.89)*** Own stocks & bonds 1.153 1.141 (1.25) (1.35) Hispanic 0.750 0.765 0.806 0.811 (3.32)*** (2.26)** (2.59)*** (1.82) Black 0.401 0.439 0.529 0.532 (11.40)*** (7.30)*** (8.68)*** (6.13)*** N 14652 8750 18665 11206
Note: a. Robust z statistics in parentheses b. ** significant at 5%; *** significant at 1% c. In each specification, control variables include age, age-squared, education, religion, father’s and mother’s education, region of residence, whether or not state recognizes common law marriage, local unemployment rate, and year dummy variables (10 in model 1, and 7 in model 2).
36
Table 5. F-Statistic for the Joint Significance of Instrumental Variables in the First Stage
Women Men Variables Model 1 Model 2 Model 1 Model 2
No. of excluded instruments 18 21 18 21 Transition from Non-partnered Own home 4.22 2.99 3.24 1.95 Have money assets 2.48 3.62 1.93 1.80 Own stocks, bonds 1.64 1.30 Transition from Cohabiting Own home 2.87 2.07 2.51 2.19 Have money assets 2.13 1.64 1.35 1.75 Own stocks, bonds 1.37 1.11 Transitions from unmarried (Non-partnered/cohabiting) Own home 5.97 4.08 5.70 3.03 Have money assets 3.00 3.90 2.16 3.02 Own stocks, bonds 1.96 1.60
Note: a. F-statistic is for a hypothesis that the instrumental variables jointly have no effect. b. Robust standard errors are calculated to account for clustering on each individual.
37
Table 6. Determinants of the Rate of Transitions Probit and IV Probit Coefficients
Women: Model 1 Women: Model 2 Men: Model 1 Men: Model 2 Variables
Probit IV Probit Probit IV Probit Probit IV Probit Probit IV Probit a. Transitions from Non-partnered to Partnered Own home -0.058 -0.349 -0.024 -1.131 0.108 0.240 0.127 0.733 (0.98) (0.75) (0.36) (1.28) (2.28)** (0.36) (2.22)** (0.77) Have money assets 0.085 -1.930 0.018 0.246 0.133 -0.649 0.074 -0.309 (2.13)** (2.76)*** (0.32) (0.27) (4.13)*** (0.87) (1.63) (0.30) Own stocks, bonds 0.105 -1.118 0.098 -2.497 (1.37) (0.89) (1.57) (1.81) N 11073 11073 6342 6342 14717 14717 8280 8280 J Statistic (Overidentification test) d 8.323 24.517 20.791 18.483 P-value for the J statistic (.939) (.139) (.187) (.424) b. Transitions from Non-partnered to Cohabiting Own home -0.033 -0.134 -0.025 0.065 0.038 0.736 0.114 -0.034 (0.38) (0.22) (0.26) (0.05) (0.58) (1.03) (1.49) (0.03) Have money assets -0.099 0.301 -0.138 1.217 -0.005 -0.714 -0.019 1.511 (1.74) (0.34) (1.68) (0.90) (0.12) (0.71) (0.32) (1.74) Own stocks, bonds -0.005 -2.074 0.006 -0.374 (0.04) (1.34) (0.07) (0.17) N 11073 11073 6342 6342 14717 14717 8280 8280 J Statistic (Overidentification test) d 11.348 19.003 15.976 22.453 P-value for the J statistic (.788) (.392) (.455) (.213)
38
Table 6 (Cont’d).
Women: Model 1 Women: Model 2 Men: Model 1 Men: Model 2
Variables Probit IV Probit Probit IV Probit Probit IV Probit Probit IV Probit
c. Transitions from Non-partnered to Married Own home -0.060 -0.426 -0.023 -1.636 0.125 -0.683 0.102 0.990 (0.91) (0.69) (0.31) (1.86) (2.33)** (0.90) (1.54) (0.75) Have money assets 0.175 -2.843 0.114 -0.549 0.204 -0.280 0.137 -1.720 (3.89)*** (3.60)*** (1.74) (0.61) (5.32)*** (0.22) (2.49)** (1.42) Own stocks, bonds 0.134 -0.678 0.124 -3.837 (1.61) (0.48) (1.79) (1.84)
N 11073 11073 6342 6342 14717 14717 8280 8280 J Statistic (Overidentification test) d 12.292 17.089 18.292 18.804
P-value for the J statistic (.724) (.517) (.050) (.404)
d. Transitions from Cohabiting to Married Own home 0.098 -1.846 0.067 -1.631 -0.001 -0.210 0.003 -0.820 (1.38) (2.54)** (0.80) (1.95) (0.02) (0.35) (0.05) (0.95) Have money assets 0.227 0.663 0.227 0.777 0.393 -0.698 0.328 -0.102 (3.89)*** (1.15) (3.11)*** (0.84) (7.16)*** (1.08) (4.95)*** (0.09) Own stocks, bonds 0.163 0.086 0.334 -1.105 (1.18) (0.06) (3.19)*** (0.44) N 3668 3668 2496 2496 4076 4076 3053 3053
J Statistic (Overidentification test) d 10.986 11.196 9.757 21.653
P-value for the J statistic (.810) (.886) (.714) (.117)
39
Table 6 (Cont’d).
Women: Model 1 Women: Model 2 Men: Model 1 Men: Model 2 Variables
Probit IV Probit Probit IV Probit Probit IV Probit Probit IV Probit
e. Transitions from Unmarried (Non-partnered/Cohabiting) to Married Own home 0.096 -0.026 0.095 -0.920 0.164 -0.471 0.158 1.002 (2.02)** (0.05) (1.74) (1.27) (3.92)*** (0.71) (3.23)*** (1.46) Have money assets 0.177 -2.435 0.151 0.598 0.250 -0.574 0.203 -0.789 (5.02)*** (2.72)*** (3.12)*** (1.06) (7.87)*** (0.66) (4.72)*** (0.98) Own stocks, bonds 0.094 -1.359 0.126 -4.358 (1.36) (1.03) (2.21)** (2.53)** N 14652 14652 8750 8750 18665 18665 11206 11206
J Statistic (Overidentification test) d 14.024 17.145 17.215 17.096
P-value for the J statistic (.597) (.513) (.190) (.517)
Note: a. z statistics in parentheses. For probit estimates, robust standard errors are calculated to account for clustering on each individual. For the IV- probit estimates, bootstrapped standard errors are reported. b. ** significant at 5%; *** significant at 1%
c. In each specification, control variables include Age, age-squared, ethnicity, education, religion, father’s and mother’s education, region of residence, whether or not state recognizes common law marriage, local unemployment rate, and year dummy variables (10 in model 1, and 7 in model 2).
d. The Hansen-Sargan J Statistic is derived from a linear estimate of the binary dependent variable model.
40
Table 7. Determinants of the Rate of Transition from Unmarried to Married:
Women and Men by Education Probit and IV Probit Coefficients in Model 1
HS or less educated More than HS educated Variables Probit IV Probit Probit IV Probit
a. Women Own house 0.106 0.546 0.085 -0.410 (1.50) (0.90) (1.32) (0.62) Have money asset 0.230 -1.999 0.122 -1.575 (5.09)*** (2.40)** (2.33)** (2.13)** First-stage F statistic Own house 2.95 4.57 Have money asset 1.79 2.95 N 7777 7777 6875 6875 b. Men Own house 0.159 -0.727 0.161 -1.033 (2.76)*** (0.97) (2.65)*** (1.72) Have money asset 0.277 0.644 0.237 -0.289 (7.28)*** (0.67) (4.40)*** (0.43) First-stage F statistic Own house 3.62 3.21 Have money asset 2.17 2.19 N 11505 11505 7160 7160
Note: a. The results reported in this table are estimated from specifications similar to model 1 in Table 6 (includes dummy indicators for homeownership and money assets). b. z statistics in parentheses. Robust standard errors are calculated to account for clustering on each individual. c. ** significant at 5%; *** significant at 1%
41
Appendix
Table A-1. Determinants of the Rate of Transitions Inclusion of Annual Income and Welfare Recipiency as Controls
Hazard ratios from Cox Proportional Hazard Model
Women Men Variables Model 1 Model 2 Model 1 Model 2 a. Transitions from Non-partnered to Partnered Own house 0.985 0.689 1.287 0.999 (0.14) (2.36)** (3.25)*** (0.01) Have money asset 1.118 1.215 1.292 1.433 (1.25) (1.39) (3.07)*** (2.99)*** Own stocks & bonds 1.237 1.056 (1.71) (0.38) Annual income 1.000 1.000 1.000 1.000 (1.64) (0.00) (2.95)*** (2.16)** Welfare recipiency 0.802 0.650 0.820 0.640 (1.90) (1.88) (1.49) (1.72) N 11067 6337 14691 8257 b. Transitions from Non-partnered to Cohabiting Own house 1.019 0.708 1.136 0.950 (0.10) (1.29) (0.90) (0.24) Have money asset 0.782 0.892 1.019 1.181 (1.66) (0.50) (0.18) (1.16) Own stocks & bonds 1.042 0.856 (0.17) (0.63) Annual income 1.000 1.000 1.000 1.000 (1.26) (0.06) (0.43) (0.44) Welfare recipiency 1.002 0.886 0.849 0.631 (0.01) (0.41) (0.93) (1.70) N 11067 6337 14691 8257 c. Transitions from Non-partnered to Married Own house 0.968 0.674 1.386 1.030 (0.26) (2.15)** (3.38)*** (0.21) Have money asset 1.321 1.434 1.532 1.694 (2.69)*** (2.13)** (4.08)*** (3.42)*** Own stocks & bonds 1.324 1.179 (1.90) (1.09) Annual income 1.000 1.000 1.000 1.000 (1.74) (0.04) (3.24)*** (2.58)*** Welfare recipiency 0.690 0.534 0.729 0.582 (2.66)*** (2.19)** (1.50) (1.39) N 11067 6337 14691 8257
42
Table A-1 (Cont’d).
Women Men
Variables Model 1 Model 2 Model 1 Model 2 d. Transitions from Cohabiting to Married Own house 1.019 1.090 0.946 0.934 (0.19) (0.71) (0.58) (0.64) Have money asset 1.324 1.351 1.759 1.620 (2.96)*** (2.50)** (6.55)*** (4.59)*** Own stocks & bonds 1.215 1.477 (1.20) (3.03)*** Annual income 1.000 1.000 1.000 1.000 (2.29)** (1.46) (9.48)*** (7.45)*** Welfare recipiency 0.908 1.066 0.616 0.584 (0.86) (0.46) (2.72)*** (2.51)** N 3665 2494 4070 3047 e. Transitions from Unmarried (Non-partnered/Cohabiting) to Married Own house 1.265 1.147 1.389 1.231 (2.82)*** (1.46) (4.55)*** (2.31)** Have money asset 1.267 1.270 1.496 1.460 (3.62)*** (2.52)** (6.63)*** (4.43)*** Own stocks & bonds 1.154 1.131 (1.26) (1.26) Annual income 1.000 1.000 1.000 1.000 (0.66) (0.50) (4.05)*** (3.18)*** Welfare recipiency 0.782 0.884 0.703 0.705 (2.79)*** (1.09) (2.46)** (1.87) N 14643 8743 18633 11177
Note: a. Robust z statistics in parentheses b. ** significant at 5%; *** significant at 1% c. In each specification, control variables include age, age-squared, ethnicity, education, religion, father’s and mother’s education, region of residence, whether or not state recognizes common law marriage, local unemployment rate, and year dummy variables (10 in model 1 and 3; 7 in model 2 and 4).
